Introduction 

The relationship between pet ownership and human health has been widely studied, with research suggesting various benefits, such as increased physical activity (Mueller et al., 2021; Utz, 2014) and psychological well-being (Herzog, 2011; McNicholas et al., 2005; Serpell, 1991). For instance, Serpell (1991) lists the health benefits of pet ownership, such as reduced headaches and difficulty concentrating, less tiredness and anxiety, etc. Mueller et al. (2021) identify key differences between pet and non-pet owners, emphasizing the link between dog ownership and higher physical activity levels. Notably, the Human-Animal Bond Research Institute (HABRI) claims that pet ownership saves Americans $23 billion annually in healthcare costs (n.d.). However, this assertion could be untenable given that the pet industry sponsors this report. Utz (2014) argues that the positive health effects of pet ownership appear to be primarily the result of selection, not increased physical activity associated with the active caretaking of pets. Mueller et al. (2021) also contend that while dog ownership was associated with more physical activity, pet ownership was not associated with overall health status if individual and family-level covariates were controlled. 

In addition, consumer behavior can be highly contextual, and pet ownership expenditure varies significantly based on various socioeconomic factors such as income and lifestyle (Applebaum et al., 2023; Mueller et al., 2021; Utz, 2014). Household income plays a significant role, with high-income households viewing pet expenses as discretionary, while lower-income households may only cover essential pet-related costs. 

This study adds to the current literature on pets’ impact on human health by exploring the spatial relationship between the two. Focusing on Toronto, the study seeks to answer the following questions: Does household spending on pets correlate with out-of-pocket healthcare expenses, and how does this relationship vary spatially? One hypothesis is that higher pet expenses may be associated with lower healthcare costs in some areas, which aligns with HABRI’s claim. However, considering the impacts of consumer behavior, an alternative hypothesis is that higher-income households, which tend to spend more on both pet care and healthcare, may drive a positive correlation between these expenses. 

Data 

Overall, this study requires data on household healthcare expenditures, pet expenses, and income to group the spatial relationship between pet spending and healthcare costs (Geographical Units: Dissemination Areas). These datasets are retrieved from SimplyAnalytics, a web-based mapping and data visualization platform. The dataset of place names in Ontario is retrieved from Ontario GeoHub. The analysis unit focuses on household population instead of general population data, as the former includes individuals living in households, excluding those in institutional or collective living settings. Since pet and healthcare expenses are primarily household-level expenditures, this selection ensures that the analysis focuses on relevant consumer groups.

During the selection process, specific datasets were preferred over their alternatives to ensure appropriate comparisons and minimize distortions in the analysis. Below is a more detailed explanation of dataset selection:  

1. Household Pet expenses 

The study uses “Household Pet Expenses, 2024 HHSpend” to capture total pet-related spending at the household level, which is the same unit for the rest of the datasets used in this study. This dataset measures pet-related financial commitment, including routine pet care, veterinary visits, discretionary spending, etc.

2. Household Current Consumption Expenditure 

This is one of the datasets for grouping populations to account for financial behaviors. “Household Current Consumption Expenditure, 2024 HHSpend” is used to categorize households based on their overall spending behavior. In particular, “current consumption” focuses on actual spending on goods and services that occur regularly at the household level. Household current consumption is the smallest economic category in the measurement of household income and expenditure which also includes pet and healthcare expenditures. Therefore, this current consumption dataset is preferred over “Average Total Expenditure,” which includes non-consumer expenditures such as investment that might introduce distortions in later comparison of pet and healthcare expenses. 

3. Healthcare Expenses 

Similarly, for healthcare costs, the selection of datasets in this study also attempts to minimize distortions. I employ “ Total Direct Costs to Household for Health Care, 2024 HHSpend,” which captures out-of-pocket spending on healthcare services. This portion of healthcare expenses excludes health insurance coverage and is, therefore, more directly comparable to pet expenses, as both represent discretionary household expenditures. It is thus preferred over the alternative dataset, “Average Total Expenditure | Health Care” which includes insurance premiums and employer or government contributions. Since they are not direct household expenses, the average total healthcare expenditure could distort the data comparison. 

Methods and Workflow 

1. Data preparation: Logarithm transformation 

As discussed, healthcare expenses, pet-related expenditures, and overall consumption patterns are closely linked to household spending habits. Higher-income households tend to spend more across all categories due to greater financial flexibility, whereas lower-income groups exhibit more constrained spending behaviors. During data cleansing, all three datasets displayed significant right skewness, indicating a strong influence from higher-income households. To address this, I applied a log transformation, replacing each variable with its natural logarithm to approximate a normal distribution. This adjustment enhances the validity of the subsequent regression analysis. Below are three sets of histograms of the datasets used in this study. On the left are three datasets (consumption, healthcare, and pet expenses) without log transformation, which are all highly right-skewed. On the right are their log-transformed counterparts. 

Fig. 1a. (left) Household Current Consumption (CAD); 1b. (right) Log Transformation of Household Current Consumption

Fig. 2a. (left) Household Healthcare (HC) Consumption (CAD); 2b. (right) Log Transformation of Household HC Expenses

Fig. 3a. (left) Household Pet Expenses (CAD); 3b. (right) Log Transformation of Household Pet Expenses

2. GLR and Spatial Autocorrelation

For spatial regression, I conduct Generalized Linear Regression (GLR) and examine spatial autocorrelation using Moran’s I. Examining spatial autocorrelation demonstrates the role of spatiality in the regression analysis of pet expenses and healthcare expenses. If Moran’s I indicates a relatively strong spatial autocorrelation, geographically weighted analysis would be conducted subsequently to take spatial relationships, such as neighboring, into consideration. 

3. Grouping and GWR Analysis of Different Consumption-Leveled Households

Then, I group households by their current consumption level (Low, Medium, High) to account for differences in spending habits. The three levels of household consumption adopted here are divided by the first and third quartiles of log-transformed data. After grouping, I conduct GWR analyses of each household group using the Model Builder in ArcGIS Pro. 

Results and Discussion 

1. GLR and Spatial Autocorrelation 

Table 1 presents the regression output from the GLR analysis examining the relationship between household healthcare (HC) expenses (dependent variable) and household consumption and pet expenses (explanatory variables). The resulting regression formula is: 

y=0.79×1+0.06×2-0.85

where y represents HC expenses, x1and x2represents consumption and pet expenses respectively. Coefficients in this formula have been rounded to 2 decimals, and all data have been log-transformed for normalization.

The coefficient for pet expenses (0.06) in the GLR model suggests a positive yet weak relationship between household pet spending and HC expenses: A 1% increase in household pet expenses is associated with a 0.06% increase in HC expenses, holding household consumption constant. Though households spending more on pets tend to have slightly higher healthcare costs, the effect size is quite small.

Variables	Coef	StdError	t_Stat	Prob
Consumption_Log	0.791751015	0.011551451	68.54126174	0
PetExpenses_Log	0.061209269	0.007396493	8.275444617	2.22045E-16
Intercept	-0.846511735	0.093999431	-9.005498523	0

Table 1. GLR of Healthcare Expenses. Dependent variable: Healthcare expenses (log-transformed); Explanatory variables: current consumption, pet expenses (both log-transformed).   

2. Spatial autocorrelation of GLR (Moran’s I) 

I conducted a Moran’s I analysis on the Std. residuals from the GLR model to assess the spatial autocorrelation in the regression results. As Table 2 shows below, the analysis reveals a highly clustered spatial pattern, with Moran’s I equal to 0.167, z-score  51.799, and p-value less than 0.0001. This result indicates that the residuals are not randomly distributed. In particular, given the extremely high z-score, there is less than a 1% likelihood that this clustering is due to random chance. This suggests that GLR may not fully account for spatial variation in the relationship between HC expenses and pet expenses, and consequently, a GWR analysis is conducted next to explore localized variations.

Moran’s Index	0.167197
Expected Index	-0.000268
Variance	0.000010
z-score	51.799220
p-value	0.000000

Table 2. Spatial Autocorrelation (Moran’s I) of GLR Std. Residuals 

3. Grouping by Household Consumption Level 

Fig. 4. Household by Current Consumption (Log-Transformed)

Consumption Group	Range Defined by Q1 & Q3	Number of Households	Percentage of Households
1 – Low Consumption	9.37 – 11.31	1270	33.98%
2 – Medium Consumption	11.31 – 11.54	1230	32.91%
3 – High Consumption	11.54 – 13.77	1238	33.12%
Total	–	3738	100%

Table 3. Household by Current Consumption (Log-Transformed) Level separated by Q1 and Q3

To account for differences in spending habits, households were grouped into three consumption levels, namely Low, Medium, and High, based on their log-transformed current consumption expenditure. The thresholds for each group were determined using the first and third quartiles (Q1 and Q3) of the log-transformed dataset. As shown in Table 3 above, households with log-transformed consumption values between 9.37 and 11.31 were classified as Low Consumption (33.98% of households), those between 11.31 and 11.54 as Medium Consumption (32.91%), and those between 11.54 and 13.77 as High Consumption (33.12%). This classification ensures that household groups are evenly distributed while reflecting meaningful differences in financial behavior. 

4. GWR Analysis of Different Consumption-Leveled Households

After grouping, separate GWR analyses were conducted for each consumption group using Model Builder in ArcGIS Pro to examine localized variations in the relationship between pet expenses and healthcare expenditures. Results are shown below. 

Fig. 5a. GWR Coefficients (Low-Consumption Households)

Fig. 5b. GWR Coefficients (Medium-Consumption Households)

Fig. 5c. GWR Coefficients (High-Consumption Households)

The GWR coefficient for pet expenses shows how pet-related spending influences healthcare expenses across space. A higher positive coefficient means a stronger positive relationship in that area, whereas a lower coefficient suggests a weaker or even negative relationship. Across different consumption levels, the majority of households have a positive relationship between pet expenses and healthcare expenses (Fig. 5a-c). This indicates that even if a household spends more on pets, they do not necessarily reduce their spending on healthcare. This pattern suggests that pet ownership and care are not seen as substitutes for healthcare expenses but may instead be complementary. Households with higher consumption levels tend to be more willing to invest in both pet and human well-being.

Notably, however, among high-consumption households (Fig. 5c), there are negative coefficients predicting the negative relationship between pet expenses and HC expenses. Areas where higher pet expenditure correlates with lower healthcare expense mainly cluster in the east and west ends of Toronto: Etobicoke and Scarborough Village. This spatial trend is significant as it may indicate that wealthier households in these areas derive greater health benefits from pet ownership, potentially reducing their need for medical services. For instance, pet companionship has been linked to lower stress levels, improved cardiovascular health, and increased physical activity (Mueller et al., 2021), which may contribute to lower healthcare expenditures in these areas. The consumption pattern on pets and healthcare in these areas aligns well with the Human Animal Bond Research Institute (HABRI)’s report, which suggests that investment in pet care tends to lead to cost savings in human healthcare in these areas. 

However, this study does not establish a clear causal relationship between pet ownership and healthcare expenditures. While the observed pattern supports the idea that pet ownership can have health benefits, the extent to which these benefits translate into reduced healthcare spending remains uncertain. Given that both pet expenditures and healthcare costs are strongly influenced by consumer behavior, their relationship is highly contingent on various socioeconomic and spatial contexts. Thus, HABRI’s contention may not be universally applicable, as factors such as income levels, access to healthcare, household size, pet type, etc, all shape this dynamic. 

Limitations and future research suggestions 

While this study identifies spatial patterns in the relationship between pet spending and healthcare expenses, future research should account for additional consumer behavior factors, such as lifestyle choices, access to healthcare, household sizes, and pet type, etc. other regression modeling method, such as multi-criteria regression could be conducted to incorporating more socioeconomic, demographic, and behavioral factors into consideration. 

In particular, Etobicoke and Scarborough Village as the only two areas where the positive relationship between pet expenses and HC expenses clusters may not be a random phenomenon. Known for being the low-income neighborhoods surrounding Toronto, households in Etobicoke and Scarborough Village may have less access to traditional healthcare services. Therefore, households might allocate more of their budget to pets for emotional support and well-being. In this sense, it might be interesting for future research to look into the correlation between pet ownership and healthcare coverage and/or accessibility for the population in these areas. Also, there might be more chances for pet owners to conduct physical activities such as dog walking in the suburbs. Future research could look into the relationship between urbanity/rurality, pet ownership, and health outcomes. 

Other spatial factors, such as access to veterinary services, green space availability, and healthcare infrastructure, could also be incorporated to further explain the geographic variation in the relationship between pet expenses and healthcare costs. Finally, temporal analyses such as a space-time cube could help determine whether pet expenditures lead to positive or negative changes in healthcare expenses over time. This would provide stronger evidence of potential causality and long-term trends in spending behavior. 

References 

Applebaum, J. W., McDonald, S. E., Widmeyer, M., Fabelo, H. E., & Cook, R. L. (2023). The impact of pet ownership on healthcare access and utilization among people with HIV. PLOS ONE, 18(11), e0292658. https://doi.org/10.1371/journal.pone.0292658

Athey, S., Blei, D., Donnelly, R., Ruiz, F., & Schmidt, T. (2018). Estimating heterogeneous consumer preferences for restaurants and travel time using mobile location data. arXiv. https://doi.org/10.48550/arXiv.1801.07826

Human-Animal Bond Research Institute. (n.d.). Healthcare cost savings. https://habri.org/health-care-cost-savings/

Henderson, S. B., Wan, V., & Kosatsky, T. (2013). Differences in heat-related mortality across four ecological regions with diverse urban, rural, and remote populations in British Columbia, Canada. Health & Place, 23, 48–53. https://doi.org/10.1016/j.healthplace.2013.04.005

Herzog, H. (2011). The impact of pets on human health and psychological well-being: Fact, fiction, or hypothesis? Current Directions in Psychological Science, 20(4), 236–239. https://doi.org/10.1177/0963721411415220

McGuirt, J. T., Wu, Q., Laska, M. N., Truesdale, K. P., Rafferty, A. P., Bell, R. A., Ammerman, A. S., & Jilcott Pitts, S. B. (n.d.). Associations between shopping patterns, dietary behaviours and geographic information system-assessed store accessibility among small food store customers. Public Health Nutrition, 25(5), 1255–1264. https://doi.org/10.1017/S1368980020005017

McNicholas, J., Gilbey, A., Rennie, A., Ahmedzai, S., Dono, J.-A., & Ormerod, E. (2005). Pet ownership and human health: A brief review of evidence and issues. BMJ, 331(7527), 1252–1254. https://doi.org/10.1136/bmj.331.7527.1252

Mueller, M. K., King, E. K., Callina, K., Dowling-Guyer, S., & McCobb, E. (2021). Demographic and contextual factors as moderators of the relationship between pet ownership and health. Health Psychology and Behavioral Medicine, 9(1), 701–723. https://doi.org/10.1080/21642850.2021.1963254

Oshan, T. M., Smith, J. P., & Fotheringham, A. S. (2020). Targeting the spatial context of obesity determinants via multiscale geographically weighted regression. International Journal of Health Geographics, 19(1), 11. https://doi.org/10.1186/s12942-020-00204-6

Serpell, J. (1991). Beneficial effects of pet ownership on some aspects of human health and behaviour. Journal of the Royal Society of Medicine, 84(12), 717–720.

Tokey, A. I., & Shioma, S. A. (2022). Spatial association between dog ownership and crime rate in New York City. Findings. https://doi.org/10.32866/001c.37094

Utz, R. L. (2014). Walking the dog: The effect of pet ownership on human health and health behaviors. Social Indicators Research, 116(2), 327–339. https://doi.org/10.1007/s11205-013-0299-6